Clinical Trial Milestones

The practice of milestone prediction in clinical trials is multifaceted, blending statistical rigor with strategic foresight. It’s about more than just adhering to a schedule; it’s about adapting to realities on the ground and ensuring that a trial can meet its objectives without wasting resources. Effective milestone management helps maintain the integrity of the trial process, ensuring that therapeutic potentials are accurately assessed while upholding the highest standards of safety and efficacy.

In clinical trial management, understanding both enrollment dynamics and event occurrence—including dropouts, cures, or any factors preventing subjects from experiencing key events—is crucial. Given the commonality of delays, with approximately 80% of trials experiencing slowdowns and about 85% failing to reach recruitment goals, the need for robust milestone prediction is evident. This prediction involves assessing practical elements such as enrollment strategies and resource allocation, which are essential to maintaining trial timelines and efficiency.

Key Challenges and Strategies in Milestone Prediction

  1. Enrollment and Event Tracking: The primary milestones in most trials involve patient enrollment and tracking event occurrences, like patient survival or endpoint achievement. In event-driven studies, such as those focusing on survival, predicting when the study might conclude or when interim analyses might be needed is paramount.

  2. Handling Practical Challenges: Addressing practical issues involves predicting enrollment timelines and identifying potential delays early. If enrollment lags, strategies might include opening new trial sites or closing underperforming ones. Proactive resource management, such as reallocating resources to more critical areas, becomes possible with accurate milestone forecasting.

  3. Data Availability and Prediction Management:

    • Data Handling: Trials might not have access to all data, particularly unblinded data. Predictors might need to rely on blinded data or assume an overall global event process rather than specific data from individual groups. The complexity of the methodology used can vary depending on data access levels.
    • Site and Subject Level Data: Access to detailed site or subject level data can provide greater flexibility and precision in predictions, allowing for more tailored adjustments in trial management.

  4. Frequency and Timing of Predictions:

    • Continuous vs. Intermittent Predictions: There is a debate between continuously updating predictions as new data comes in and waiting for patterns to develop. Continuous updates might disrupt the trial’s natural progression, particularly if early trial phases naturally exhibit slower recruitment.
    • Resource Management: Overly frequent adjustments might lead to inefficient resource use, such as unnecessary expansion of trial sites which can overwhelm staff and inflate costs.

  5. Special Considerations for Survival Studies:

    • Delayed Effects: In studies involving treatments like immunotherapies, delayed effects are common, where the treatment’s impact takes time to manifest. This must be factored into milestone predictions to avoid premature conclusions about treatment efficacy or participant response.

  6. External Predictions: Employing external experts for milestone predictions can reduce bias and provide access to a broader range of methodologies. External predictors, less influenced by internal trial dynamics, might offer a clearer, unbiased perspective.

Additional Considerations

  • Safety and Regulatory Benchmarks: Besides primary outcomes, secondary considerations might include safety analyses and regulatory compliance milestones. These are crucial for maintaining ethical standards and satisfying regulatory requirements.
  • Sponsor-Specific Requests: Tailoring milestone predictions to meet specific sponsor requests or interests from regulatory bodies can also guide the frequency and method of prediction updates.

Enrollment & Event Milestone Predictions

Enrollment Prediction

Enrollment prediction is a crucial aspect of clinical trial planning and management, serving as a foundational metric for assessing a trial’s timeline and resource allocation. It encompasses predicting both the rate and completeness of participant recruitment over the course of the study. This process not only impacts the financial and logistical aspects of a trial but also its scientific validity, as timely enrollment ensures that the trial can achieve its intended statistical power and objectives.

Implementing effective enrollment predictions requires a multi-faceted approach: - Data integration: Combining data from multiple sources, including historical trial data, current site performance, and external factors. - Continuous monitoring: Regularly updating predictions based on new data to stay responsive to changing conditions. - Stakeholder communication: Using prediction data to maintain open dialogue with sponsors and adjust expectations and strategies as needed.

Initial and Mid-Trial Predictions

Enrollment predictions typically begin with estimating how long it will take to recruit the full sample size needed to meet the study’s power requirements. This involves assessing: - Demographics: The availability and willingness of the target population to participate. - Competing studies: Other ongoing trials that could affect participant availability. - Site capabilities: Each site’s ability to recruit and manage participants.

Mid-trial predictions evaluate whether enrollment is on track to meet planned timelines. Adjustments might be needed if the trial is progressing faster or slower than expected.

Challenges with Early and Later Phase Trials

  • Early-phase trials often struggle with recruitment due to the experimental nature of the treatments and the typically smaller pool of eligible participants.
  • Later-phase trials may face competition from established treatments, making it harder to recruit participants unless the new treatment offers clear advantages.

Site-Specific Predictions

More sophisticated approaches to enrollment prediction involve modeling each recruitment site or region separately. This allows for: - Detailed tracking: Identifying which sites are underperforming. - Resource reallocation: Shifting resources to more effective sites or boosting those that are lagging. - Adaptive strategies: Adjusting recruitment tactics based on real-time data.

Methodologies for Enrollment Prediction

Simple Statistical Models These include linear or polynomial models that provide a basic forecast based on past recruitment rates.

Piecewise Parametric Models These models identify changes in recruitment pace, such as an initial slow start followed by a faster rate, allowing for more nuanced predictions.

Simulation-Based Modeling Simulation offers a flexible and dynamic approach to modeling recruitment. It allows for:

  • Scenario testing: Simulating different recruitment strategies to see potential outcomes.
  • Bootstrapping: Using resampling techniques to estimate prediction intervals and assess uncertainty.

Bayesian Models These incorporate prior data and expert opinions to refine predictions, adapting as new data becomes available during the trial.

Machine Learning Approaches While not covered in detail here, machine learning methods can analyze complex datasets to predict recruitment outcomes, potentially uncovering hidden patterns that affect enrollment.

Event Prediction

In survival trials, the occurrence of key events such as death or disease progression is fundamental to determining the trial’s timeline and outcomes. The predictive modeling of these events is complex due to the multifaceted nature of survival data, which can include various competing risks and time-dependent factors.

Event-Driven Endpoints: In many clinical trials, especially those concerning life-threatening conditions, the trial’s endpoint is driven by the accumulation of specific events among participants (e.g., death, disease progression). The number of events directly impacts the trial’s power and its ability to provide statistically meaningful results. Without a sufficient number of events, the trial cannot conclude or make robust inferences.

Challenges in Event Prediction

  1. Continuous Enrollment: If enrollment is ongoing, predictions must account not only for current participants but also for how new enrollees might alter the event dynamics.
  2. Competing Risks: Factors such as alternative treatments, dropouts, or other medical interventions can influence the timing and occurrence of the primary events of interest.
  3. Event Scarcity: In scenarios where events are fewer than expected, it can delay the trial significantly, affecting timelines and potentially increasing costs and resource usage.

Modeling Techniques for Event Prediction

  • Parametric Models: These models, such as the exponential or Weibull models, assume a specific distribution for the time until an event occurs. They are straightforward but often too simplistic for complex survival data.

  • Piecewise Parametric Models: These improve on simple parametric models by allowing different parameters in different phases of the study, accommodating varying hazard rates across the trial’s duration.

  • Simulation-Based Models: Simulations provide a flexible and dynamic approach to understanding how different factors might impact event rates. This method is particularly useful in survival trials where complex interactions between patient characteristics and treatment effects need to be considered.

Practical Implementation of Event Prediction

  • Exponential Models: Assume constant hazard rates throughout the trial period. This is simplistic but can serve as a baseline for understanding baseline event rates.

  • Piecewise Exponential Models: Offer more flexibility by dividing the trial into segments, each with its own hazard rate, better modeling the natural progression of disease or treatment effects over time.

  • Two-Parameter Models: These models account for the duration a participant has been in the study, adjusting the event probability based on this tenure. They are useful in long-term studies where the risk of an event may increase or decrease over time.

  • Model Selection and Evaluation: Employing information criteria like AIC (Akaike Information Criterion) helps in selecting the best-fitting model amongst various candidates. Advanced techniques might also dynamically allocate change points to adapt the model to observed data patterns more accurately.

Reference

Background

  • Lamberti, M.J. (2012). State of Clinical Trials Industry. Thomson Centerwatch, Clinical Trials Arena. Retrieved from https://www.clinicaltrialsarena.com/analysis/featureclinical-trial-patient-recruitment
  • McDonald, A.M., Knight, R.C., Campbell, M.K., Entwistle, V.A., Grant, A.M., Cook, J.A., Elbourne, D.R., Francis, D., Garcia, J., Roberts, I., & Snowdon, C. (2006). What influences recruitment to randomised controlled trials? A review of trials funded by two UK funding agencies. Trials, 7(1), 1-8.
  • Bower, P., Brueton, V., Gamble, C., Treweek, S., Smith, C.T., Young, B., & Williamson, P. (2014). Interventions to improve recruitment and retention in clinical trials: a survey and workshop to assess current practice and future priorities. Trials, 15(1), 1-9.
  • Cognizant. (2015). Patient Recruitment Forecast in Clinical Trials. Retrieved from https://www.cognizant.com/whitepapers/patients-recruitment-forecast-in-clinical-trials-codex1382.pdf
  • Gkioni, E., Rius, R., Dodd, S., & Gamble, C. (2019). A systematic review describes models for recruitment prediction at the design stage of a clinical trial. Journal of Clinical Epidemiology, 115, 141-149.
  • Kearney, A., Harman, N.L., Rosala-Hallas, A., Beecher, C., Blazeby, J.M., Bower, P., … Gamble, C. (2018). Development of an online resource for recruitment research in clinical trials to organise and map current literature. Clinical Trials, 15(6), 533-542. https://doi.org/10.1177/1740774518796156

Enrollment

  • Lee, Y.J. (1983). Interim recruitment goals in clinical trials. Journal of Chronic Diseases, 36(5), 379-389.
  • Comfort, S. (2013). Improving clinical trial enrollment forecasts using SORM. Applied Clinical Trials, 22(5), 32.
  • Carter, R.E., Sonne, S.C., & Brady, K.T. (2005). Practical considerations for estimating clinical trial accrual periods: application to a multi-center effectiveness study. BMC Medical Research Methodology, 5(1), 1-5.
  • Carter, R.E. (2004). Application of stochastic processes to participant recruitment in clinical trials. Controlled Clinical Trials, 25(5), 429-436.
  • Senn, S. (1998). Some controversies in planning and analyzing multi‐centre trials. Statistics in Medicine, 17(15-16), 1753-1765.
  • Anisimov, V.V., & Fedorov, V.V. (2007). Modelling, prediction and adaptive adjustment of recruitment in multicentre trials. Statistics in Medicine, 26(27), 4958-4975.
  • Anisimov, V. (2009). Predictive modelling of recruitment and drug supply in multicenter clinical trials. In Proc. of Joint Statistical Meeting (pp. 1248-1259).
  • Anisimov, V.V. (2011). Statistical modeling of clinical trials (recruitment and randomization). Communications in Statistics - Theory and Methods, 40(19-20), 3684-3699.
  • Bakhshi, A., Senn, S., & Phillips, A. (2013). Some issues in predicting patient recruitment in multi‐centre clinical trials. Statistics in Medicine, 32(30), 5458-5468.
  • Jiang, Y., Guarino, P., Ma, S., Simon, S., Mayo, M.S., Raghavan, R., & Gajewski, B.J. (2016). Bayesian accrual prediction for interim review of clinical studies: open source R package and smartphone application. Trials, 17(1), 1-8.
  • Gajewski, B.J., Simon, S.D., & Carlson, S.E. (2008). Predicting accrual in clinical trials with Bayesian posterior predictive distributions. Statistics in Medicine, 27(13), 2328-2340.
  • Abbas, I., Rovira, J., & Casanovas, J. (2007). Clinical trial optimization: Monte Carlo simulation Markov model for planning clinical trials recruitment. Contemporary Clinical Trials, 28(3), 220-231.
  • Moussa, M.A.A. (1984). Planning a clinical trial with allowance for cost and patient recruitment rate. Computer Programs in Biomedicine, 18(3), 173-179.

Events

  • Lakatos, E. (1988). Sample sizes based on the log-rank statistic in complex clinical trials. Biometrics, 229-241.
  • Fang, L., & Su, Z. (2011). A hybrid approach to predicting events in clinical trials with time-to-event outcomes. Contemporary Clinical Trials, 32(5), 755-759.
  • Rufibach, K. (2016). Event projection: quantify uncertainty and manage expectations of broader teams. Retrieved from http://bbs.ceb-institute.org/wp-content/uploads/2016/06/Kaspar-event_tracking.pdf
  • Walke, R. (2010). Example for a Piecewise Constant Hazard Data Simulation in R. Max Planck Institute for Demographic Research.
  • Goodman, M.S., Li, Y., & Tiwari, R.C. (2011). Detecting multiple change points in piecewise constant hazard functions. Journal of Applied Statistics, 38(11), 2523-2532.
  • Guyot, P., Ades, A.E., Beasley, M., Lueza, B., Pignon, J.P., & Welton, N.J. (2017). Extrapolation of survival curves from cancer trials using external information. Medical Decision Making, 37(4), 353-366.
  • Royston, P. (2012). Tools to simulate realistic censored survival-time distributions. The Stata Journal, 12(4), 639-654.
  • Crowther, M.J., & Lambert, P.C. (2012). Simulating complex survival data. The Stata Journal, 12(4), 674-687.